Offer Selection Using Sequential Selection Operations

1. A method of selecting one or more offers to be served to a respondent from a set of offers having multiple characteristics, wherein each characteristic has more than one possible value, at least one characteristic corresponding to a location, the method being performed using one or more processors in a computing system, wherein the method of selecting comprises: at a hardware-based website host server operating at least one processor of the processors serving one or more offers to a respondent device operating a processor; collecting response data relating to the offers at the website host server and transmitting the response data to a hardware-based decision server operating at least one processor of the processors; receiving at the decision server, from the website host server, a request to score some or all of the offers against a set of criteria; in response to the request: performing in real time a series of at least three successive scoring operations, wherein: each scoring operation of the series of successive scoring operations is performed by a decision engine executed by the decision server and comprises determining respective scores for different possible values of one of the multiple characteristics, three or more decision engines forming a linked decision engine; each of the at least three scoring operations is carried out according to a respective reinforcement machine learning model including one or more variables and modified by response data, wherein each scoring operation comprises selecting the value for one of the multiple characteristics having the highest score and wherein the selected value is said result of at least one previous scoring operation used as a variable in one or more subsequent scoring operations; the location characteristic determining a location on a computer for the offer; the one or more variables of the reinforcement machine learning model for at least a second scoring operation and any scoring operations subsequent to the second operation include scores for all characteristics scored in previous scoring operations; and a machine learning model for at least a second scoring operation and any scoring operations subsequent to the second operation scoring an additional characteristic not scored in previous scoring operations; using the determined scores to select in real time one or more offers from the set of offers, the one or more offers being in the form of a web banner in a web page or a web page; and serving the selected one or more offers to the respondent device.

2. The method of claim 1 further comprising calculating the sum of the scores of selected values from respective scoring operations to derive a total score for the series of scoring operations.

3. The method of claim 1 wherein said set of criteria comprises values of one or more variables describing the respondent and at least some of the variables describing the respondent are included in at least some of the respective reinforcement machine learning models used in the scoring operations.

4. The method of claim 1 comprising performing at least one subsequent series of said scoring operations in which the order of carrying out the operations is different from the order of carrying out the operations in the previously performed series of scoring operations.

5. The method of claim 4 further comprising calculating the sum of the scores of selected values from respective scoring operations to derive a total score for each series of scoring operations.

6. The method of claim 5 comprising selecting one or more offers to be served to a respondent having the values of characteristics selected in the highest scoring series of scoring operations.

7. A computing system for use in performing a selection of one or more offers among a set of offers having multiple characteristics, comprising: a hardware-based website host server operating at least one processor serving one or more offers to a respondent device operating a processor, the one or more offers being in the form of a web banner in a web page or a web page, the website host server collecting response data relating to the offers, and transmitting the response data to a hardware-based decision server operating at least one processor, the decision server receiving at from the website host server a request to score offers against a set of criteria; one or more processors programmed to perform in real time a series of at least three selection operations each resulting in the selection of one offer by implementing a plurality of reinforcement machine learning-based decision engines, three or more reinforcement machine learning-based decision engines forming a linked reinforcement machine learning-based decision engine wherein each decision engine performs a scoring operation on a characteristic and selects the value of one characteristic having the highest score and wherein the selected value is said result of at least one previous scoring operation used in one or more subsequent scoring operations; the location characteristic determining a location on a computer for the offer, wherein each reinforcement machine learning-based decision engine is programmed to make a selection from among a plurality of offers based on a plurality of input variables, and wherein for at least one of the plurality of reinforcement machine learning-based decision engines the input variables include scores for all characteristics scored in previous scoring operations of decision engines and a reinforcement machine learning-based decision engine scoring an additional characteristic not scored in a previous scoring operation, at least one characteristic corresponding to a location; the website host server serving the selected offer to the respondent device.

8. A computing system according to claim 7 in which each reinforcement machine learning-based decision engine is programmed to operate according to a respective mathematical model in order to make said selection.

9. A computing system according to claim 8 wherein each model includes a plurality of variables and at least some of the variables are used by all of the models.

10. A computing system according to claim 9 in which the model for each reinforcement machine learning-based decision engine includes as a variable a selection made by each of the other decision engines.

11. A computing system according to claim 10 in which the one or more processors are programmed such that the reinforcement machine learning-based decision engines may operate in any order and any reinforcement machine learning-based decision engine may take into account any previously made decision or selection by another reinforcement machine learning-based decision engine in response to the same selection request.

12. A method of selecting a combination of two or more offers from a plurality of offers having multiple characteristics to be served to potential respondents, the method being performed using one or more processors in a computing system, wherein the method of selecting comprises: at a hardware-based website host server operating at least one processor of the processors serving one or more offers to a respondent device operating a processor; collecting response data relating to the offers at the website host server and transmitting the response data to a hardware-based decision server operating at least one processor of the processors; receiving at the decision server, from the website host server, a request for a selection of a combination of two or more offers; in response to the request: performing in real time a series of at least three selection operations each resulting in the selection of one offer, each of the at least three selection operations being carried out according to a respective reinforcement machine learning model executed by a decision engine executed by the decision server and including one or more variables and modified by response data, each decision engine performing a scoring operation comprising selecting the value for one of the multiple characteristics having the highest score and wherein the selected value is said result of at least one previous scoring operation used as a variable in one or more subsequent scoring operations; the location characteristic determining a location on a computer for the offer; three or more decision engines forming a linked decision engine, wherein the variables of the reinforcement machine learning model for the second and any subsequent selection operation include scores for all characteristics scored in previous selection operations, and a reinforcement machine learning model for at least a second selection operation and any selection operations subsequent to the second operation scoring an additional characteristic not scored in previous selection operations, at least one characteristic corresponding to a location; outputting a signal identifying the selected offers; and serving the selected offers to the respondent device.

13. The method of claim 12 wherein the offers of the plurality of offers are organized in respective sets, each selection operation is performed on a different one of the sets, and the combination of two or more offers comprises offers selected from different sets.

14. The method of claim 13 wherein the request includes values for one or more variables which are included in the respective reinforcement machine learning models for all of the selection operations.

15. The method of claim 14 wherein the variables included in the reinforcement machine learning model comprise characteristics of a particular respondent to whom the combination of two or more offers is to be served.

16. The method of claim 15 comprising performing at least one further series of selection operations, wherein each further series of selection operations is performed in a different order from any previously performed series of selection operations, and determining a score for each selected offer in each selection operation.